blixt

10 ms·

121.
by blixt 2y ago
Yeah I did also find it to be mostly accurate. However, seeing the sources I felt like I kind of have to check all the languages just in case it picked up information from a random "X ways to do Y" article that might not have been
122.
by blixt 2y ago
They add the same Option+Space (on macOS) shortcut that ChatGPT has, but the app feels a bit like a quick build overall, as it's just a web container. Hopefully they'll add some nice features that wouldn't work in a browser s
123.
by blixt 2y ago
What I really hope this helps solve is covering for the huge lag in knowledge cutoff. A recent example is where it went "oh you're using Go 1.23 which doesn't exist so that's clearly the problem in your Dockerfile, let m
124.
by blixt 2y ago
One thing that is quite unfortunate with the state of SEO and the web in general today is that when I asked "what are the latest versions of common programming languages and when were they released?" a large amount of the sources
125.
by blixt 2y ago
I have the opposite experience with Chat and Compose. I use the latter much more. The "intelligence level" of the Chat is pretty poor and like the author says, starts with pointless code blocks, and you often end up with AI slop a
126.
by blixt 2y ago
Ah gotcha, yeah I was always a bit confused about the whole thing but that makes sense. I've only used SWC directly as a WASM bundle to do in-browser realtime compilation and analysis of .tsx files, which works great, but that was pre-
127.
by blixt 2y ago
It is easy to get JS build fatigue. But I love the fact that these competitors push the incumbents forward. SWC was one of the earlier Rust builders, which then became absorbed into Vercel and turned into Turbopack (is my understanding), an
128.
by blixt 2y ago
Yeah I think CRDT or OT can work well for multiplayer rich text editing for example, but then I would scope the domain to just that in such a way that you still keep the bulk of your document "dumb", e.g. as JSON patches on a big
129.
by blixt 2y ago
I'm completely into the idea that your data should be document-centric, this I've seen working at a very large scale (i.e. 100+ MB JSON blobs being updated at 60 FPS with a large team of people editing at the same time) with the s
130.
by blixt 2y ago
Sorry, I meant bitwise operators, such as: ~ >> << >>> | &
131.
by blixt 2y ago
JS numbers technically have 53 bits for integers (mantissa) but all binary operators turns it into a 32-bit signed integer. Maybe this is related somehow to the setTimeout limitation. JavaScript also has the >>> unsigned bit shift
132.
by blixt 2y ago
I'm constantly impressed by the design of DOs. I think it's easy to have a knee-jerk reaction that something is wrong with doing it this way, but in reality I think this is exactly how a lot of real products are implicitly structu
133.
by blixt 2y ago
I saw the reference to “apps like Figma” and as one of the people that worked on Framer’s (also a canvas based app) database which is also local+multiplayer I find it hard to imagine how to effectively synchronize canvas data with a relatio
134.
by blixt 2y ago
I was impressed by Microsoft’s AICI where the idea is a WASM program can choose the next tokens. And relatedly their Guidance[1] framework which can use CFGs and programs for local inference to even speed it up with context aware token fill
135.
by blixt 2y ago
A friend of mine made this in-browser neural network engine that could run millions of multi-layer NNs in a simulated world at hundreds of updates per second and each network could reproduce and evolve. It worked in the sense that the netwo
136.
by blixt 2y ago
Maybe the validator should be something like Object.getPrototypeOf(document.createElement("a")).constructor !== HTMLUnknownElement I noticed several functional HTML elements (like "font") were not in the list, so might b
137.
by blixt 2y ago
Channels require a different goroutine to send values while also receiving them (which is how you’d have two loops communicating, essentially what you get from range funcs). There’s nothing stopping you from doing this but it does mean you
138.
by blixt 2y ago
Very cool! Good to see the shoutout to Dennis' Holograph.so work as well. I played around with that one to make some fun things: A game of cat and mouse: https://x.com/blixt/status/1797384954172625302 An anal
139.
by blixt 2y ago
From Kyutai, demo'd today: https://www.youtube.com/live/hm2IJSKcYvo
140.
Live demo of Moshi, audio-to-audio LLM with 160 ms latency(us.moshi.chat)
1 points by blixt 2y ago|1 comments
141.
by blixt 2y ago
I was also exploring randomness in JS at some point and found lots of interesting things! One was that the Alea PRNG algorithm[1] by Johannes Baagøe performed faster on JavaScript's floating point numbers. Another was that Dieharder[2]
142.
by blixt 2y ago
It’s fair in some cases, but indeed - not in others - e.g., code completion with LLMs is a time saver and worth paying for, compared to any other tech out there. At least this is how I see it, I know some people say it introduces mistakes b
143.
by blixt 2y ago
I meant more that the % is so low that even if the usage is on top of all other usage (not a completely clear statement to make), it's like starting to mention any other thing in the long tail of technology leaving behind a "massi
144.
by blixt 2y ago
> It costs $700,000 per day in energy costs to run ChatGPT 3.5, according to recent estimates, and leaves behind a massive carbon footprint in the process. Compared to what? I wouldn't defend LLMs as "worth their electricity&qu
145.
by blixt 2y ago
You could take a look at a project I recently made, it’s very easy to put together and run a server to do what you want! https://github.com/blixt/sol-mate-eink
146.
by blixt 2y ago
Probably Daniel Gross picked it up from Nat Friedman? 1. Nat Friedman has this site: https://nat.org/ 2. They made this together: https://nfdg.com/ 3. And then this: https://andromeda.ai/ 4
147.
by blixt 2y ago
Kind of sounds like OpenAI when it started, so will history repeat itself? Nonetheless, excited to see what comes out of it.
148.
by blixt 2y ago
They’ve been explicit that only some knowledge related requests get sent to OpenAI, and only after the user has been asked “ok to send this to ChatGPT?” If they allowed your private email contents or any other private context to leak outsid
149.
by blixt 2y ago
Did I miss the explanation of how they trained their image generation models? It's brave of a company serving creative professionals to generate creative works with AI. I'm a fan of using generative AI, but I would have expected t
150.
by blixt 2y ago
Yeah this is Dall-E incapable of cleanly writing text. The 2 C is just a “hallucination” from the 26 C. This is a hobby project and as such I prioritized the text looking unique every time (and usually more artistic than in that image), rat

More ›